Andrea E. Martin, PhD
Bitte klicken Sie hier, um das vorgesehene Video von Vimeo zu laden.
Dabei werden Daten an Vimeo übermittelt — beachten Sie diesbezüglich unsere Datenschutzerklärung
Human language is a fundamental biological signal with computational properties that differ from other perception-action systems: hierarchical relationships between words, phrases, and sentences, and the unbounded ability to combine smaller units into larger ones, resulting in a "discrete infinity" of expressions. These properties have long made language hard to account for from a biological systems perspective and within models of cognition. One way to begin to reconcile the language faculty with both these domains is to hypothesize that, when hierarchical linguistic representation became an efficient solution to a computational problem posed to the organism, the brain repurposed an available neurobiological subroutine. Under such an account, a single mechanism must have the capacity to perform multiple, functionally-related computations, e.g., detect and represent the linguistic signal, and perform other cognitive functions, while, ideally, oscillating like the human brain. We show that a well-supported symbolic-connectionist model of analogy (Discovery Of Relations by Analogy; Doumas, Hummel, & Sandhofer, 2008) oscillates while processing sentences - despite being built for an entirely different purpose (learning relational concepts and performing analogical reasoning). The model processes hierarchical representations of sentences, and while doing so, it exhibits oscillatory patterns of activation that closely resemble the human cortical response to the same stimuli (cf. Ding, Melloni, Zhang, Tian, & Poeppel, 2016). From the model, we derive an explicit computational mechanism for how the brain could convert perceptual features into hierarchical representations across multiple timescales, providing a linking hypothesis between linguistic and cortical computation. We argue that this computational mechanism – using time to encode hierarchy across a layered network, while preserving (de)compositionality – can satisfy the computational requirements of language, in addition to performing other cognitive functions. Our results suggest a formal and mechanistic alignment between representational structure building and cortical oscillations that has broad implications for discovering the first principles of linguistic computation in the human brain.